Data Engineer vs Data Scientist: Roles, Skills & Hiring Differences

A VP of Data at a mid-market SaaS company in Costa Mesa called us about eighteen months ago. She had one open headcount, board pressure to “do something with AI,” and a senior data scientist she’d hired six months earlier who was, in her words, “not producing anything.” We pulled the scientist’s calendar with her permission. The woman had spent the entire previous quarter writing dbt models and arguing with the data warehouse team about ingestion schedules. She had a PhD in statistics. She had not built a model since her interview. Six months. No model. Just plumbing work she never signed up for.

The fix was a data engineer. Two of them. Wrong hire, wrong order, and the cost was a year. Once the warehouse was real, the scientist shipped her first revenue-impact model in nine weeks. She wasn’t the problem. The order was.

The plain version of the data engineer vs data scientist question is this. A data engineer builds and maintains the systems that move data from where it lives to where it gets used. A data scientist takes that prepared data and uses statistics, modeling, and machine learning to answer questions a business cares about. Both roles touch Python. Both touch SQL. Almost nothing else about the day-to-day work overlaps, and pretending it does is the single most common mistake I see in JDs that hit our queue.

I’m Devin Hornick, one of the founding partners at KORE1. We’ve placed engineers, scientists, and analytics talent at companies across California and nationally for two decades, and the data side of our practice has grown faster than any other vertical over the last five years. This guide is the conversation I have with hiring managers when they call us with one open headcount and the wrong instincts about how to spend it. I’ll flag where we benefit from you hiring through us and where we genuinely don’t. There are spots you should not call us at all, and I’ll point those out too.

We help clients staff both sides of the data org through our data science and data engineering staffing practice, which is the relevant context for everything below. Read it knowing where I’m coming from. Bias on the table.

The Plain-English Definition

A data engineer designs and maintains the pipelines, warehouses, and infrastructure that turn raw operational data into clean, queryable datasets. A data scientist uses those datasets to answer business questions through statistics, experimentation, and machine learning models. Engineers ship infrastructure. Scientists ship insights. Different artifacts. Different success metrics. Different humans.

That’s the version that wins on Google. It’s also the version that has gotten roughly half the hiring managers I’ve talked to in the past year into trouble. Every educational article on the SERP says “engineers build pipelines, scientists build models,” and then leaves you holding a job description with no idea which one your company actually needs first. Worse, none of them mention the third role most data orgs need before either of the first two. That’s the analytics engineer, and we’ll get to her.

Here’s the part the career-advice articles skip. The boundary between these two roles has been moving for about four years. A senior data engineer in 2026 builds in dbt, owns the semantic layer, writes Python that looks more like a backend engineer’s, and increasingly handles the production machine learning infrastructure that used to belong to a separate ML platform team. A senior data scientist in 2026 spends less time on Jupyter notebooks and more time on causal inference, experimentation design, and explaining to product leaders why the A/B test result they want is statistically meaningless. The roles have specialized as the field has matured. Anyone telling you a data scientist is just “an analyst with Python” hasn’t been in a real data org since 2019.

Side-by-Side: What Each Role Actually Owns

Here’s the comparison that I wish existed when our clients are writing the JD. The career articles get the columns right. They get the row that matters wrong, which is the “where it fails” row.

The “where the hire fails” row is the one nobody else publishes. It’s also the one that drives 80% of the failed data placements I’ve seen. Read it twice.

The Skills That Actually Matter (and the Ones on the JD That Don’t)

JDs for these roles are notoriously bad. Almost universally bad. We rewrite about 60% of the data JDs that come into our queue before we’ll send candidates against them, and the rewrites are usually load-bearing because the original JD would have attracted exactly the wrong half of the market. The standard pattern is a giant skills bullet list with everything from Python to Tableau to Kubernetes to “experience with stakeholders.” That kind of JD attracts nobody good and offends everybody great.

What a real data engineer does in week one

Drops into your warehouse. Reads your existing dbt project, if you have one. If you don’t, asks where the source-of-truth tables live and discovers that there are six “source of truth” tables that disagree with each other. Picks one to start fixing. Writes a small audit query that compares revenue across them. Sends a Slack message that begins with “so this is interesting.” Within two weeks she has a refactored ingestion pattern, a draft data quality framework, and an opinion about your warehouse spend that will make your CFO want to talk to her.

The real skill list, written like a hiring manager would actually screen against it? SQL at a level that makes the rest of your team uncomfortable. Python good enough to write a custom Airflow operator without Googling it. dbt at the level where she has opinions about exposures and metricflow. Snowflake, BigQuery, or Databricks at the level where she can read a query plan and tell you why it’s slow. That’s the real list. Not the twenty-five-bullet JD list. Maybe seven things, all of them deep.

What a real data scientist does in week one

Asks for the question. Then asks for the question again, because the first answer was a metric, not a question. Then she asks about the data, and someone tells her it’s “in the warehouse,” and she discovers that “in the warehouse” means a 240-column raw events table that nobody has documented since 2022. She files a ticket with the data engineering team. Then she does what she can with what’s there, ships a quick analysis with caveats, and starts building trust by being honest about what the data can and can’t say.

Real skills? Causal inference. Experimentation design that survives contact with a product manager. Statistics deep enough to spot when a Bayesian framing would save the analysis. Python, of course, but the meaningful Python is in libraries like statsmodels and PyMC, not pandas tricks. And the soft skill that gets undervalued every single time: the ability to tell a senior leader that the answer they want is not in the data, without making it feel like an attack.

Both roles list “Python” on the resume. The Python is completely different. A data engineer’s Python looks like a backend service. A data scientist’s Python looks like a research notebook with a few production-grade exports. Neither one would pass a code review on the other’s repo without complaining the entire time.

Compensation Reality in 2026

Salary data for these roles has been moving fast and the public ranges lag the market. Here’s what we’re actually placing at, cross-referenced against the Bureau of Labor Statistics OOH page for data scientists, Levels.fyi for top-tier comp, and our own placement queue from the last six months. Ranges are US base, not total comp. Add 15-30% for total comp at funded startups and large tech.

1. Junior data engineer: $95K to $125K. Almost always one to three years of experience, frequently a recent CS or data grad with a bootcamp specialization.
2. Mid-level data engineer, three to five years, $130K to $160K. This is the largest band by volume. If you’re hiring one DE and one DS, this is usually where the engineer lives.
3. Senior data engineer: $160K to $195K base. Add a meaningful equity package at any company that calls itself a startup.
4. Staff and principal DE, top tier, $210K to $280K base. The handful of people who can architect a real data platform from scratch and also have the political skills to get a CFO to fund it. A scarce talent pool. Dozens, not hundreds. Levels.fyi has good public data here for the FAANG-tier numbers if you want to validate.
5. Junior data scientist: $100K to $130K. PhDs sometimes start higher, sometimes lower. Yes, the variance is that wide.
6. Mid-level DS, $135K to $170K, three to six years and at least one production model.
7. Senior data scientist: $165K to $210K base.
8. Staff and principal DS, top tier, $220K to $310K base. ML-heavy or causal-inference specialists are the high end of this band.

Here’s the inversion that surprises hiring managers. Five years ago, a senior data scientist out-earned a senior data engineer in almost every market. That gap has closed. In some California markets it has flipped. Senior DEs at a Snowflake-stack company are now matching or exceeding what the equivalent DS makes, because the supply of true senior DEs is even thinner than the supply of senior DSs. Don’t write a comp band based on what you remember from 2021, because the floor has moved fifteen to twenty percent in three years and the ceiling has moved more than that for the senior end of the engineering side. Stale comp bands are the second-biggest reason searches stall on our end.

For deeper salary detail and how the bands shift by metro, see our 2026 data engineer salary guide and the live ranges in the KORE1 salary benchmark assistant. Both pull from current placement data, not stale aggregator averages.

Which One Should You Hire First?

This is the only question that matters if you have one open headcount. Just one.

The diagnostic is straightforward. Walk into the office of whoever currently runs reporting and ask them, “If I asked you for our churn rate by customer segment for last quarter, broken out by acquisition channel, how long would it take?” If the answer is more than four hours, your problem is data engineering. Hire the engineer. Do not hire a scientist yet, no matter how loudly the CEO is saying the word “AI.”

If the answer is “I’d have it before lunch,” your data infrastructure works. Now the question becomes what kind of question your business actually needs answered. Predictive forecasting and experimentation? That’s a data scientist. Production model deployment and serving? That’s an ML engineer (a different role we’ll touch on in the FAQ). Better dashboards, cleaner metrics, and a semantic layer the rest of the org can self-serve? That’s an analytics engineer, and almost no companies hire her at the right time.

I mentioned a logistics client earlier. The rest of what happened. They came to us in early 2025 wanting a data scientist. We pushed back hard and recommended two data engineers first. They were skeptical but they trusted the relationship. Six weeks after the second DE started, the warehouse was clean enough that the eventual data scientist hire shipped a route-optimization model in her first sprint. Total time from first conversation to ROI: about four months. The version where they hired the scientist first would have taken eighteen months and probably ended in a quiet termination.

Now flip the scenario. A health-tech company in San Diego came to us last fall convinced they needed an ML engineer. They had a working warehouse, two analysts, and a PM screaming for “predictive insights.” We talked through their actual use case for two hours. What they needed was an analytics engineer to build the metrics layer their PM was guessing at, plus a part-time DS contractor for one well-scoped predictive analysis. Total spend was under half what an ML engineer would have cost them. We placed the analytics engineer through our analytics staffing practice in three weeks. The ML engineer they thought they needed was, frankly, an aspiration their data wasn’t ready for.

The Org Chart Question Nobody Asks

Where these roles sit on the org chart matters more than the job title. A lot more.

Data engineers usually report into engineering. Sometimes a dedicated VP of Data, sometimes the VP of Engineering directly, sometimes a platform team lead. Their work is infrastructure. They live next to backend engineers in code review and deployment cadence, and they should. A data engineer who reports into Marketing is a data engineer who’s about to leave. Watch for it.

Data scientists are the opposite. They work best when they’re embedded close to the question, which usually means Product, Strategy, Finance, or a dedicated Analytics function. A data scientist who reports into Engineering will spend most of her time being asked to write production code she didn’t sign up for, and within a year she will either quit or quietly transition into being a worse version of one of the senior backend engineers on her team.

The mistake we see most often is putting both roles under the same first-line manager, usually a director who came up through one side or the other and quietly favors that side in performance reviews. A manager who has actually done both sides of the work is the only kind who survives running both teams under one roof, and if you don’t have that person, you need two different managers. No shortcut.

Common Hiring Mistakes We See

Hiring a scientist before the warehouse exists. The Costa Mesa story from the opener is not unusual. We see it about once a quarter. The pattern is always the same. Board pressure for AI, a CEO who has read one McKinsey article, and a head of HR who Googles “data scientist salary” and writes a JD. Eight months later, the scientist is either gone or has quietly become a data engineer with a more expensive title.

Treating data engineering as a junior version of data science. They are not the same career track. An engineer is not “a scientist who hasn’t gotten her PhD yet.” They have completely different aptitudes, different academic backgrounds (DEs more often come from CS or software engineering, DSs more often from stats, math, physics, or economics), and different growth paths. Treating one as a stepping stone to the other insults both, and we have watched companies lose strong DEs because the head of data made an offhand comment about “growing them into the science side eventually” during a one-on-one.

The “data unicorn” JD. Everyone in hiring has read it. It asks for ten years of Spark, a PhD in statistics, production ML deployment experience, dashboarding expertise, and stakeholder management at the C-suite level. That person exists. A few hundred of them work in the US, and none of the ones we know are answering recruiter messages this quarter. Pick a role. Just one. Stop trying to clone a unicorn that the market spent five years not producing in any volume.

Skipping the analytics engineer. If your company has decent operational data and your real question is “why are our numbers wrong and inconsistent across teams,” your hire is an analytics engineer. Not a DE, not a DS. Almost no JDs we see correctly identify this need on the first try. Maybe one in twenty.

When KORE1 Is and Isn’t the Right Fit

The honest version, since I promised one.

We’re a strong fit when you’re hiring senior or staff-level data talent on a real timeline, when you need contract-to-hire flexibility (about a third of our data placements start as C2H), or when you’ve burned a search internally and want someone who knows the California and national talent map. We’re particularly useful for the in-between metro hiring problems: a company in Irvine that needs someone willing to be hybrid two days a week and the local pool is shallow. That kind of metro-specific search is most of our queue.

We are not the right fit for very junior data hires. Use university programs and internship pipelines. We’re also not the right fit for academic or research-track ML roles with pure-research output expectations. Those roles want academic networks we don’t pretend to have. And if you already have a strong internal recruiting function placing data talent at your target seniority, you don’t need us. You can also see our full data engineer hiring guide for the DIY version of the search process. It’s free and reasonably honest about where it falls short.

Want a second opinion on which role to open first? Talk to a KORE1 recruiter and we’ll walk through your specific situation on a thirty-minute intake call. The intake call costs nothing and most of the time we end up either pointing you at the right role or telling you to keep looking on your own.

Common Questions

So which one earns more in 2026?

$165K to $210K is the senior base range for both, and the gap is closer to a wash than the old career articles suggest. Senior data scientists still edge out senior DEs at the top of the market, but the floor for senior DE has come up faster than the floor for senior DS over the last two years. In some California markets we’re now placing senior DEs above senior DSs at companies on a Snowflake or Databricks stack.

Can a data engineer become a data scientist? (And should they?)

Sometimes. Not often. The skills don’t transfer the way people assume. A great data engineer who tries to retrain into data science usually ends up as a mediocre data scientist who’s overqualified for the engineering work she left behind. The better move is usually to deepen on the engineering side, move into platform or ML infrastructure, and let her data instincts inform the architecture. We’ve seen maybe a dozen successful DE-to-DS transitions in two decades. We’ve seen many more attempts.

Both roles, or just one of them, for a 150-person company?

Wrong question, slightly. The right question is what the next twelve months of your business demand. If the answer is “we need to know things we don’t currently know about our customers and our operations,” that’s a data scientist (assuming the data is queryable). If it’s “we need our reports to be trustworthy and faster,” that’s a data engineer or an analytics engineer. Most companies under 200 people need one before they need the other. Companies over 500 usually need both. Sequence matters more than the headcount itself, and getting the sequence wrong is the single most expensive data hiring mistake we see at companies in the 80-to-300 employee range.

How long does a senior data engineer search take?

Forty to seventy days for a well-scoped search at a fair comp band. We’ve closed senior DE searches in three weeks when the JD was tight and the client moved fast. We’ve also seen senior DE searches drag past four months when the comp band was twenty percent under market and the hiring manager wouldn’t budge. The biggest single predictor is how much the client has actually decided what they want before the search starts. Half of all delays are JD revisions. Half. Not a third, not a quarter, half. The other half is split between comp negotiations and the client losing momentum because a board meeting got rescheduled.

Data scientist versus ML engineer, does the gap actually matter?

Different jobs. Different muscles. A data scientist designs and validates models. An ML engineer takes a model and turns it into a production service that handles real traffic, monitors drift, retrains on schedule, and doesn’t fall over at 3 AM. The skills overlap is maybe 30%. Most companies under 100 employees do not need a dedicated ML engineer; they need a data scientist who can hand off cleanly to a backend engineer for productionization. We staff both through our AI/ML engineer staffing practice, and the conversation about which role you need usually takes about an hour on intake.

Is the data scientist job market still strong in 2026?

Yes for senior, choppy for junior, and that split is sharper than it was two years ago. The 2022 era of every Series B opening five DS reqs at once is over. What replaced it is a more discriminating market where companies want senior DSs with domain depth, specific industry experience, and a track record of shipping models that affected revenue. The junior DS market is harder than it’s been in five years, partly because LLM tooling has absorbed a lot of the “first-pass analysis” work that used to be entry-level. Senior demand is as strong as ever. Maybe stronger. We just placed three senior DSs in the first quarter of 2026 alone, all into companies that had explicitly told us six months earlier they were “pausing data science hiring.” That pause is over for the senior end of the market.